For nearly two decades, a coronary artery calcium score of zero has meant one thing to patients and their doctors: relief, Zero calcium, low risk, go home and keep living your life. It is one of the most reassuring numbers in preventive medicine. New research using AI powered imaging is now complicating that reassurance, not by throwing it out, but by showing that zero was never quite as complete a story as it sounded.For nearly two decades, a coronary artery calcium score of zero has meant one thing to patients and their doctors: relief, Zero calcium, low risk, go home and keep living your life. It is one of the most reassuring numbers in preventive medicine. New research using AI powered imaging is now complicating that reassurance, not by throwing it out, but by showing that zero was never quite as complete a story as it sounded.
Why zero became the number everyone trusted
Coronary artery calcium scoring, sometimes called a CAC scan or a heart scan, measures calcified plaque buildup in the coronary arteries using a non contrast CT scan. The result is expressed as an Agatston score, and the categories are simple enough that patients remember them. A score of zero means no detectable plaque and a low risk of heart attack. Scores climbing from the low hundreds into the thousands mark progressively higher risk, up to a score above 1,000, which research has identified as a distinct, very high risk category on its own.
The evidence behind zero is genuinely strong. Long running research from the Multi Ethnic Study of Atherosclerosis, known as MESA, along with the Rotterdam Study, found that a calcium score of zero predicts a meaningfully lower rate of coronary events over ten years, outperforming even polygenic genetic risk scores in some comparisons. A 2025 presentation at the American College of Cardiology’s annual meeting, led by Jeffrey Anderson at Intermountain Health, followed patients for more than two years and found coronary events occurred in only 0.12 percent of people under 65 and 0.25 percent of those over 65 with a zero score, a two to three fold lower event rate compared with people who had any detectable calcium at all. A 2026 review published in Frontiers in Radiology also described this as the power of zero, and called it one of the most reproducible, well validated tools in preventive cardiology.
None of that has been overturned. What has changed is the growing recognition that zero calcium is not the same thing as zero coronary disease, and AI powered imaging is exactly what is making that distinction visible for the first time.
The gap zero was never built to see
Calcium scoring only measures calcified plaque. It says nothing about soft, non calcified plaque, which can be just as dangerous and, in some cases, more prone to rupture. Because standard CAC scoring cannot see it at all, a patient can have significant coronary artery disease and still walk away with a perfect zero.
New data presented at the Society of Cardiovascular Computed Tomography’s 2026 annual meeting put a hard number on that gap for the first time. Research using HeartFlow’s AI powered Plaque Analysis platform, presented at SCCT 2026, found that more than 50 percent of symptomatic patients with a zero coronary artery calcium score actually harbored significant non calcified plaque that traditional CAC scoring could not detect at all. That is not a small edge case. It means over half of a specific, symptomatic patient population walked away with the most reassuring possible test result while carrying real, undetected coronary disease.
HeartFlow’s technology works by applying AI to coronary CT angiography scans to quantify total plaque volume, separating calcified from non calcified plaque and estimating blood flow impact through a technology called CT FFR, without requiring an invasive catheter based procedure. A separate SCCT 2026 study tested whether HeartFlow’s plaque quantification held up consistently as hospitals transition from older CT scanner technology to newer photon counting CT systems, analyzing more than 12,000 patients across eight clinical sites and finding no significant differences in plaque volume, stenosis grading, or blood flow measurements between the two scanner types, an important reliability check as the underlying hardware itself changes.
A new scanner that makes even smaller plaque visible
The hardware shift matters here in its own right. Photon counting CT, a newer generation of CT scanner technology, is moving from research settings into routine clinical use, and it directly affects how calcium scoring works. Cynthia McCollough, a medical physicist at Mayo Clinic and former president of the American Association of Physicists in Medicine, explained in reporting from Cardiovascular Business that photon counting CT’s improved spatial resolution allows it to detect much smaller coronary calcifications than older scanners ever could. Her framing of the stakes was blunt: an Agatston score of zero on an older scanner might become a non zero score entirely once a patient is imaged on a photon counting system, because the difference between zero and not zero is, in her words, the difference between having coronary artery disease and not having it.
Separate research published in JACC in 2026, using micro CT validation in cadaveric arteries, found photon counting detector CT actually reduces calcium overestimation compared with older energy integrating detector CT, meaning the newer technology is not simply more sensitive in a noisy way, it appears to be more accurate in both directions. And in a Harvard Medical School and Brigham and Women’s Hospital study of asymptomatic patients with extremely high existing calcium scores above 1,000, photon counting coronary CT angiography was able to confirm which of those high risk patients did not actually have functionally significant blockages, using CT based blood flow estimates that all fell in a safe range, potentially sparing some very high calcium score patients from unnecessary invasive procedures.
Put together, photon counting CT is reshaping calcium scoring from both directions at once. It can turn some real zeros into non zeros by catching plaque older scanners missed, and it can add reassurance for some very high scores by confirming blood flow isn’t actually restricted. Neither of those things was possible with the technology hospitals have relied on for the past two decades.
Inflammation: the risk factor calcium scoring cannot measure at all
A second and separate line of research presented at SCCT 2026 pushes the same point from a different angle. Work discussed by Kenneth Chan and covered in Diagnostic Imaging examined the role of coronary inflammation in identifying elevated cardiovascular risk specifically in patients with zero or low calcium scores, essentially asking whether a biological process calcium scoring cannot see at all might explain some of the risk hiding inside a reassuring score. A related presentation at SCCT 2026, the FORTIFY study led by Todd Villines, examined whether coronary inflammation, measured through something called the Fat Attenuation Index or FAI Score, could reliably be extracted from routine coronary CT scans, including scans taken without contrast dye, expanding how widely this kind of inflammation analysis could realistically be deployed. Coverage of the same research area by TCTMD summarized the core finding plainly: quantifying coronary inflammation adds meaningful predictive value to cardiovascular risk stratification beyond what calcium score and plaque volume alone can provide, even in scans that were never designed to measure inflammation in the first place.
An AI tool that beat a standard blood test, using the same zero score scans
Perhaps the most striking evidence in this entire area comes from a study using the MESA cohort itself, the same long running dataset that helped establish the power of zero in the first place. Researchers developed a system called AI CAC, which applies AI directly to standard coronary calcium scan images, the exact same scans radiologists already read for calcium scoring, to extract an entirely different signal: structural measurements of the heart itself, including left ventricular size, that are associated with future heart failure risk.
The results were directly compared against NT proBNP, a well established blood biomarker doctors already use to help predict heart failure. In statistical models predicting heart failure, adding AI CAC’s imaging derived measurements to ten known clinical risk factors outperformed adding NT proBNP to those same risk factors, and the improvement held up even when limited specifically to patients with a coronary artery calcium score of zero. In other words, the exact scan that told a zero score patient they were low risk for coronary events already contained, hidden inside the same images, a separate signal about heart failure risk that only became visible once AI was applied to look for it.
The nuance that keeps this honest
None of this means a zero calcium score is meaningless, and it is worth being precise about that, because overstating this finding would be its own kind of harm. A 2025 review published in the journal Cureus, titled bluntly Is Zero Truly Zero, examined exactly this question and concluded that while calcium scoring remains a valuable and largely reliable prognostic tool, certain patients, particularly those with specific ethnic backgrounds or confounding risk factors, can still carry meaningful cardiovascular risk despite a zero score, and that these subgroups deserve more individualized attention rather than a blanket zero equals safe interpretation. The 2026 Frontiers review reached a similar balance, reaffirming that the power of zero remains real and well supported for near term risk de escalation in most patients, while also noting that advances in AI, spectral CT, and standardized reporting are specifically expanding what counts as an accurate zero in the first place.
The honest synthesis is this. For most people, a zero calcium score is still exactly as reassuring as it has always been. For a meaningful subset of patients, particularly those who are symptomatic despite a zero score, or who carry specific inflammatory or structural risk factors invisible to calcium alone, AI powered imaging is now able to show that zero was incomplete, not wrong.
A separate but related discovery, hiding in a routine EKG
This same pattern, a familiar test containing a hidden signal that only AI reveals, shows up again in a completely different corner of cardiology. Researchers at Michigan Medicine developed an AI model that can diagnose coronary microvascular dysfunction, a condition affecting the heart’s smallest blood vessels that standard cardiac tests routinely miss entirely, using nothing more than a routine ten second EKG strip, the same simple test performed in nearly every doctor’s office and emergency room. Coronary microvascular dysfunction disproportionately affects patients, often women, who are told their hearts are healthy based on standard testing despite real, underlying disease. Like the calcium scoring research above, this is not a new test. It is AI finding a diagnosis that was sitting, unread, inside data clinicians were already collecting.
The tools themselves, and where they stand with regulators
As with mammography, it helps to know that cardiac AI imaging is not one product but a competitive field, and knowing the names matters if you are evaluating what your own hospital or imaging center actually offers.
HeartFlow’s Plaque Analysis and its related CT FFR technology are already FDA cleared and have been used clinically for several years, with the SCCT 2026 data specifically aimed at reinforcing confidence in the platform as it moves across scanner generations. Rafael Rubinshtein, a cardiologist who has spoken publicly about cardiac CT technology at SCCT meetings, has described CT FFR as one of the key technologies helping move cardiac CT beyond simple diagnosis and toward more comprehensive, ongoing patient management, specifically because it can estimate the physiological impact of a blockage without an invasive catheter procedure. That distinction, estimating physiological impact rather than only anatomical appearance, is part of why CT FFR has continued gaining traction as more clinical evidence accumulates behind it.
Photon counting CT scanners themselves represent a bigger institutional decision than a single software product, since adopting one means a hospital is replacing core imaging hardware, not just installing new analysis software on top of an existing scanner. Mayo Clinic already runs photon counting CT as a routine workhorse system rather than a specialty tool reserved for unusual cases, though McCollough has noted that many health systems will likely start by reserving the technology for high value applications at a central hospital before expanding it further, given that photon counting scanners currently carry a higher price tag than conventional CT systems.
AI CAC, the tool that outperformed NT proBNP for heart failure prediction, is at an earlier stage than HeartFlow or photon counting CT. The research described above comes from the MESA research cohort rather than a commercial product with its own FDA clearance, which means it represents strong evidence for a future clinical tool rather than something available at your local imaging center today. That gap, between what research has proven is possible and what has actually been cleared, packaged, and deployed at scale, is a pattern worth remembering: it is the same gap that shaped the regulatory debate we covered in our companion post on AI in mammography, where evidence from a large randomized trial arguably outpaced what some FDA labels currently authorize.
If you have had a calcium score come back as zero, this research is not a reason for alarm, and it should not be read as one. The overwhelming evidence still supports that a zero score meaningfully lowers your risk. What this research adds is a more precise picture for specific situations: if you are symptomatic despite a zero or low score, if you have risk factors your doctor has flagged as unusual for your score, or if your imaging center has access to newer photon counting CT or AI plaque analysis tools, it is entirely reasonable to ask your cardiologist whether additional evaluation, such as coronary CT angiography with plaque quantification, makes sense in your specific case. This is exactly the kind of conversation AI is meant to prompt, not replace.
The larger pattern
This is the same pattern that shows up across AI in healthcare more broadly, which is the whole argument of our pillar guide on the myth of healthcare dependence on AI. AI is not quietly running cardiology while doctors watch. It is revealing signals inside tests that have existed for decades, that human eyes and older technology genuinely could not see, while leaving the final clinical judgment, and the decision about what to do next, exactly where it has always been i.e “doctors”.
Possible Concern:
It means your risk is meaningfully lower than someone with a detectable score, supported by large, long running studies including MESA and the Rotterdam Study. It does not guarantee zero risk. Research presented at SCCT 2026 found more than half of symptomatic patients with a zero score had significant non calcified plaque that calcium scoring cannot detect, and a 2025 review in Cureus found certain patient subgroups can carry real risk despite a zero score.
Photon counting CT is a newer generation of CT scanner technology with improved spatial resolution that can detect smaller coronary calcifications than older scanners. According to Mayo Clinic physicist Cynthia McCollough, this means a calcium score of zero measured on an older scanner could become a non zero score on a photon counting scanner, since the technology is simply able to see plaque that previous scanners missed entirely.
AI CAC is an AI system that analyzes the same coronary calcium scan images radiologists already review, but extracts a different signal, structural heart measurements linked to future heart failure risk, rather than just the calcium score itself. In research using the MESA cohort, adding AI CAC to standard risk factors outperformed adding the blood biomarker NT proBNP for predicting heart failure, including in patients with a calcium score of zero.
Yes, in specific documented cases. Researchers at Michigan Medicine developed an AI model that diagnoses coronary microvascular dysfunction, a condition standard cardiac testing often misses, using only a routine ten second EKG. Separately, research on coronary inflammation presented at SCCT 2026, including the FORTIFY study, found that inflammation measurements extracted from routine CT scans add cardiovascular risk information beyond what calcium scoring alone provides.
For most people with a zero score and no symptoms, current guidance does not call for additional imaging. If you are symptomatic despite a zero score, or your doctor has flagged other risk factors, research now supports that a conversation about additional evaluation, such as coronary CT angiography with AI powered plaque analysis, is reasonable and increasingly available.
